A5100704226- Video Surveillance and Tracking Methods
- Anomaly Detection Techniques and Applications
- Machine Learning and ELM
- Human Pose and Action Recognition
- Image and Signal Denoising Methods
- Advanced Manufacturing and Logistics Optimization
- Neural Networks and Applications
- Fire Detection and Safety Systems
- Robotics and Sensor-Based Localization
- Advanced Adaptive Filtering Techniques
- Domain Adaptation and Few-Shot Learning
- Scheduling and Optimization Algorithms
- Visual Attention and Saliency Detection
- Generative Adversarial Networks and Image Synthesis
- Advanced Image and Video Retrieval Techniques
- Assembly Line Balancing Optimization
- Multimodal Machine Learning Applications
- Advanced Battery Technologies Research
- Image Enhancement Techniques
- Advanced Image Fusion Techniques
- Image Processing Techniques and Applications
- Data Stream Mining Techniques
- Advanced Image Processing Techniques
- Forensic Anthropology and Bioarchaeology Studies
- Dental Radiography and Imaging
Xi'an Jiaotong University
2016-2025
University of Malaya
2025
East China Normal University
2013-2024
Xinjiang University
2024
Baidu (China)
2024
Fuzhou University
2021-2024
National University of Defense Technology
2024
Shanghai University
2023
Nanjing University of Science and Technology
2023
Capital Medical University
2022
Automatic analysis of highly crowded people has attracted extensive attention from computer vision research. Previous approaches for crowd counting have already achieved promising performance across various benchmarks. However, to deal with the real situation, we hope model run as fast possible while keeping accuracy. In this paper, propose a compact convolutional neural network which learns more efficient small number parameters. With three parallel filters executing operation on input...
In real-world crowd counting applications, the densities in an image vary greatly. When facing density variation, humans tend to locate and count targets low-density regions, reason number high-density regions. We observe that CNN focus on local information correlation using a fixed-size convolution kernel Transformer could effectively extract semantic by global self-attention mechanism. Thus, estimate crowds accurately while it is hard properly perceive On contrary, has high reliability but...
Abstract Recently, Siamese‐based trackers have drawn amounts of attention in visual tracking field because their excellent performance. However, object on Unmanned Aerial Vehicles platform encounters difficulties under circumstances such as small objects and similar interference. Most existing methods for aerial adopt deeper networks or inefficient policies to promote performance, but most can hardly meet real‐time requirements mobile platforms with limited computing resources. Thus, this...
Nowadays, the acoustic detection is widely used for defect diagnosis of gas insulated substations (GIS) in normal operation and factory tests. In this paper order to develop a data analyzer system make an assistant diagnosis, characteristic signals generated by different artificial defects such as protrusions, floating shield, void spacer bouncing particles are investigated. Some meaningful parameters behind detected extracted discussed, which distinguish background noise, partial discharge...
Crowd counting, i.e., estimation number of the pedestrian in crowd images, is emerging as an important research problem with public security applications. A key component for counting systems construction models which are robust to various scenarios under facts such camera perspective and physical barriers. In this paper, we present adaptive scenario discovery framework counting. The system structured two parallel pathways that trained different sizes receptive field represent scales...
Predicting the trajectory of pedestrians in crowd scenarios is indispensable self-driving or autonomous mobile robot field because estimating future locations around beneficial for policy decision to avoid collision. It a challenging issue humans have different walking motions, and interactions between objects current environment, especially themselves, are complex. Previous researchers focused on how model human-human but neglected relative importance interactions. To address this issue,...
Owing to their universal approximation capability and online learning manner, kernel adaptive filters have been widely used in nonlinear systems modeling. Under Gaussian assumption, traditional algorithms utilize the well-known mean square error(MSE) as a cost function get optimal solutions. For non-Gaussian situations, MSE will not properly represent statistics of error, hence degrade performance. In recent years, an information theoretic learning(ITL) based criterion called Maximum...
Image steganalysis, detecting hidden data in digital images, is essential for enhancing security. Traditional steganalysis methods typically rely on large, pre-labeled image datasets, which are difficult and costly to compile. To address this, this paper introduces an innovative approach that combines active learning off-policy Deep Reinforcement Learning (DRL) improve with minimal labeled data. Active allows the model intelligently choose unlabeled images should be annotated, thus reducing...
To investigate the effect of thermal characteristics a motorized spindle system on precision machine tool, error model for axial expansion and radial declination is proposed. With CNC coordinate boring as an object, using five-point method to calibrate errors by eddy current sensors elongation tilted values, temperatures measurement points are obtained PT100. The relationships between rotational speed temperature field, analyzed. Then fuzzy clustering analysis used group optimize variables,...
Crowd counting models in highly congested areas confront two main challenges: weak localization ability and difficulty differentiating between foreground background, leading to inaccurate estimations. The reason is that objects are normally small high-level features extracted by convolutional neural networks less discriminative represent objects. To address these problems, we propose a learning framework for crowd counting, which composed of masked feature prediction module (MPM) supervised...
We propose two synthetic aperture radar (SAR) complex image compression schemes based on DLWT_IQ and DLWT_FFT. encodes the real parts imaginary of images using directional lifting wavelet transform (DLWT) bit plane encoder (BPE), while DLWT_FFT converted by fast Fourier (FFT). Compared with discrete transform-IQ (DWT_IQ), improves peak signal-to-noise ratio (PSNR) up to 1.28 dB reduces mean phase error (MPE) 21.74%; compared DWT_FFT, PSNR 1.22 MPE 20.32%. Moreover, proposed increase 3.34...
Abstract Background Oral Squamous Cell Carcinoma (OSCC) is one of the most severe cancers in world, and its early detection crucial for saving patients. There an inevitable necessity to develop automatic noninvasive OSCC diagnosis approach identify malignant tissues on Optical Coherence Tomography (OCT) images. Methods This study presents a novel Multi‐Level Deep Residual Learning (MDRL) network benign(normal) from OCT images trains 460 captured 37 The diagnostic performances are compared...
In integrated circuit manufacturing industry, in order to meet the high demand of electronic products, wafers are designed be smaller and smaller, which makes automatic wafer defect detection a great challenge. The existing methods mainly based on precise segmentation one single wafer, relies high-cost complicated hardware instruments. performance obtained is unstable because there too many limitations brought by implementations such as camera location, light source product location. To...